Medical AI company OpenEvidence has completed a $250 million Series D funding round, achieving a valuation of $12 billion—double its previous valuation of approximately $6 billion. The round was co-led by Thrive Capital and DST, with participation from GV, Sequoia, Kleiner Perkins, and Coatue. The company offers clinical search and Q&A tools based on medical literature, claiming coverage across over 10,000 hospitals and medical centers in the U.S., with about 40% of American physicians using its platform. In December 2025, it recorded approximately 18 million monthly clinical consultations.

OpenAI CFO Sarah Friar revealed that the company's compute capacity and revenue are closely correlated: from 2023 to 2025, compute scale increased from 0.2GW to 1.9GW, while revenue grew from around $2 billion to over $20 billion. The report also notes that OpenAI remains in a high-investment phase, with estimated 2025 expenses of approximately $22 billion and revenue of about $9 billion, projecting profitability by 2030. Its consumer market share is said to have declined from around 90% in 2024 to 60–70% in 2025.

Databricks has released and open-sourced BlackIce, a containerized 'red teaming toolkit' for AI security testing, bundling 14 commonly used tools into a single reproducible environment to reduce enterprise dependency and configuration costs when building evaluation setups. The toolkit covers areas such as prompt injection, data leakage, hallucination detection, and supply chain security, mapping capabilities to MITRE ATLAS and the Databricks AI Security Framework (DASF), enabling teams to perform comprehensive validation and compliance alignment under standardized frameworks.

Yelp announced the acquisition of AI lead management platform Hatch for $270 million in cash, aiming to strengthen its customer acquisition and SaaS capabilities for local businesses. Hatch focuses on residential service industries (such as HVAC, plumbing, and electrical), offering automated tools for lead assignment, follow-up automation, and conversion optimization to help businesses respond faster to customer inquiries and improve closing rates. Yelp plans to integrate Hatch’s technology into its advertising and merchant services ecosystem to enhance coverage and monetization efficiency in high-value verticals.

Z.ai has launched the GLM-4.7-Flash large model, featuring a 30B-A3B Mixture-of-Experts (MoE) architecture and a context window of 131,072 tokens. It introduces a 'Preserved Thinking mode' to enhance contextual coherence in multi-turn agent tasks. Public benchmarks include SWE-bench Verified 59.2, AIME 25 91.6, GPQA 75.2, τ²-Bench 79.5, and LCB v6 64.0. The model is accessible via Z.ai API and web demo, and supports local deployment within the Hugging Face Transformers ecosystem.

Ollama has announced support for the Anthropic Messages API, allowing developers to retain their Claude Code workflows while shifting underlying inference to open-weight models (e.g., from Meta, Google, Mistral, Alibaba) running on local or private servers—reducing reliance on a single cloud provider. This approach enables interoperability without embedding proprietary Claude model weights, improving cost efficiency and data control. The update does not open-source Claude Code nor alter the proprietary nature of Claude models, but provides a more standardized integration path for localized agent toolchains.

Meta CTO Andrew Bosworth stated that the Superintelligence Labs has delivered its first AI models internally, roughly six months after formation, marking the team’s transition into the delivery phase. Specific model names and capabilities were not disclosed, though prior speculation suggests potential inclusion of a text model codenamed Avocado and an image/video model codenamed Mango, possibly advancing toward external release in Q1. The team is described as primarily focused on supporting internal operations (recommendations, ad ranking, Meta AI) rather than direct cloud-based monetization.

The European Data Protection Board (EDPB) and the European Data Protection Supervisor (EDPS) issued a joint opinion on the European Commission’s proposal to streamline implementation of the AI Act, expressing support for balancing consistent enforcement with regulatory burden reduction. However, they emphasized that simplification must not undermine accountability or fundamental rights protection. Regulators oppose relaxing core obligations such as registration of high-risk AI systems and remain cautious about expanding the use of sensitive personal data without stronger safeguards. They support innovation-enabling measures like regulatory sandboxes but call for clear governance boundaries and supervisory roles.

Cisco announced expanded compatibility support for Intel Gaudi 3 AI accelerators on its Nexus 9000 series switches, targeting enterprises and cloud customers building training and inference clusters. The certified configuration includes the Nexus 9364E-SG2 switch and OSFP-800G-DR8 optical modules, providing 800GbE non-blocking connectivity for servers equipped with eight Gaudi 3 accelerators, and supporting RoCEv2 transport. Gaudi 3 servers offer 24 200GbE ports for accelerator interconnects while preserving external network and management interfaces. Cisco Nexus Dashboard provides RoCE visibility and automation deployment templates.